Presents the work done to im.rove the recognition rate in an isolated word recognition problem.with single utterance training. The negative effect of errors (due to insufficient training data) in estim.ted m.del param...
ISBN:
(纸本)0769507506
Presents the work done to im.rove the recognition rate in an isolated word recognition problem.with single utterance training. The negative effect of errors (due to insufficient training data) in estim.ted m.del param.ters is com.ensated by fusing the inform.tion obtained from.Hm. evaluation and those generated for the word length and voicing at the beginning and end of the word. A Bayesian causal tree structure is developed to accom.lish the fusion. The final decision is m.de on one of the three candidates which are m.st likely according to Hm. evaluation. The reliability of the Hm. ordering is im.roved by applying variance flooring.
In this paper, we com.are several feature sets based on the param.tric and non-param.tric representations of speech. Param.tric representations are reflection coefficients, LPC derived cepstral coefficients (CCs) and ...
详细信息
In this paper, we com.are several feature sets based on the param.tric and non-param.tric representations of speech. Param.tric representations are reflection coefficients, LPC derived cepstral coefficients (CCs) and line spectral frequencies (LSFs). Non-param.tric representations are based on m.l-frequency cepstral coefficients (m.CCs). These different representations are evaluated by their scores of recognition, for a speaker independent, isolated word recognizer based on hidden m.rkov m.dels (Hm.s).< >
In this paper, a fixed point variable bit-rate (VBR) m.xed excitation linear predictive coding (m.LP/sup Tm.) vocoder is presented. The VBR-m.LP vocoder is also im.lem.nted on a Tm.320C54x and it achieves virtually in...
详细信息
In this paper, a fixed point variable bit-rate (VBR) m.xed excitation linear predictive coding (m.LP/sup Tm.) vocoder is presented. The VBR-m.LP vocoder is also im.lem.nted on a Tm.320C54x and it achieves virtually indistinguishable federal standard m.LP quality at bit-rates between 1.0 to 1.6 kb/s. The backbone of VBR-m.LP vocoder is sim.lar to that of federal standard m.LP. It utilizes a novel sub-band based voice activity detector in the back-end of encoder to discrim.nate background noise from.speech activity. Since proposed detector uses only param.ters extracted in the encoder, its com.utational com.lexity is very low.
This paper exam.nes the effect of the m.m.nt-m.tched single Gaussian approxim.tion, which is m.de in various m.ltiple-m.del filtering applications to approxim.te a Gaussian m.xture, on the Bayesian filter perform.nce....
详细信息
This paper exam.nes the effect of the m.m.nt-m.tched single Gaussian approxim.tion, which is m.de in various m.ltiple-m.del filtering applications to approxim.te a Gaussian m.xture, on the Bayesian filter perform.nce. The estim.tion error caused by the approxim.tion is analysed for both the prediction and the m.asurem.nt updates of a Bayesian filter. An approxim.te form.la is found for the covariance of the error caused by the approxim.tion for a general Gaussian m.xture with arbitrary com.onents. The calculated error covariance is used for obtaining a m.xed m.ltiple-m.del estim.tion algorithm.which has a perform.nce near that of GPB2 with less com.utations.
Naturalness of synthetic speech depends on appropriate m.deling of prosodic aspects. m.stly, three prosody com.onents are m.deled: segm.ntal duration, pitch contour and intensity. In this study, we present our work on...
详细信息
ISBN:
(纸本)2951740816
Naturalness of synthetic speech depends on appropriate m.deling of prosodic aspects. m.stly, three prosody com.onents are m.deled: segm.ntal duration, pitch contour and intensity. In this study, we present our work on m.deling segm.ntal duration in Turkish by using m.chine-learning algorithm.. The m.dels predict phone durations based on attributes such as phone identity, neighboring phone identities, lexical stress, position of syllable in word, part-of-speech inform.tion, word length in num.er of syllables and position of word in utterance. Obtained m.dels predict segm.nt durations better than m.an duration approxim.tions.
This paper introduces the use of genetics-based algorithm.in the reduction of 24 param.ter set (i.e the base set) to a 5,6,7,8 or 10 param.ter set, for each speaker in text-independent speaker identification. The feat...
详细信息
This paper introduces the use of genetics-based algorithm.in the reduction of 24 param.ter set (i.e the base set) to a 5,6,7,8 or 10 param.ter set, for each speaker in text-independent speaker identification. The feature selection is done by finding the best features that discrim.nates a person from.his/her two closest neighbors. The experim.ntal results show that there is approxim.tely 5% increase in the recognition rate when the reduced set of param.ters are used. Also the am.unt of calculation necessary for speaker recognition using the reduced set of features is m.ch less than the am.unt of calculation required using the com.lete feature set in the testing phase. Hence it is m.re desirable to use the subset of the com.lete feature set found using the genetic algorithm.suggested.
A m.them.tical m.del is developed to sim.late the hydraulic behaviour and the biodegradation process in a landfill site. The m.del incorporates governing equations that describe processes influencing the leachate prod...
详细信息
ISBN:
(纸本)0921303912
A m.them.tical m.del is developed to sim.late the hydraulic behaviour and the biodegradation process in a landfill site. The m.del incorporates governing equations that describe processes influencing the leachate production and the biochem.cal processes taking place during the stabilisation of organic wastes, including leachate flow, hydrolysis, acidogenic, and m.thanogenesis. The m.del considers the significant changes in m.isture distribution and leachate quality that m.y occur during the operational period of the landfill. The m.del results are com.ared with the leachate production, chloride concentration, BOD, pH, and m.thane production data from.Keele Valley Landfill in Ontario, Canada. Encouraging agreem.nt has been observed between the real landfill data and the m.del results.
暂无评论